Joumal of Fluorine Chemistry, 14 (1979) 115-129 0 Elsevier Sequoia S.A., Lausanne -Printed in the Netherlands

Received: February 19,1979

INORGANIC VOLATILE FLUORIDES OBTAINED FROM ELECTRICAL

DECOMPOSITION OF SULFUR HEXAFLUORIDE IN A OUARTZ TUBE

115

GIOVANNI BRUNO, PI0 CAPEZZUTO AND FRANCESCO CRAMAROSSA

Centro di Studio per la Chimica dei Plasmi de1 C.N.R.

Istituto di Chimica Generale ed Inorganica dell'Universita

Via Amendola 173, 70126 - BARI. ITALY

ABSTRACT

Recent investigations on sulfur hexafluoride decomposition

have shown the need of a rapid and efficient method for the

qualitative and quantitative analysis of the reaction products.

An analytical method for characterizing the gas mixture obtai-

ned from the decomposition of sulfur hexafluoride in a quartz

reactor submitted to an r.f. discharge, is presented. A combi-

nation of gas-chromatographic, mass spectrometric and infrared

spectrophotometric techniques has shown the presence of SF6,

S02F2, SOF4, SOF2, SiF4 and F2 in the gas mixtures examined.

For quantitative

found to be most

INTRODUCTION

purposes a gas-chromatographic method has been

suitable.

The growing interest devoted to the study of sulfur hexa-

fluoride decomposition is not only due to its peculiar die-

lectric properties '[l], for which a wide utilization as electric

insulating gas is made, but also to its potential use as a good

116

source of atomic fluoride for the etching of semiconductors

in the integrated circuits industry [2] and for the produc-

tion, in electrical discharge, of vibrationally excited hydro-

gen fluoride chemical lasers.

More recently laser induced decomposition of SF6 has been

made in connection with the isotope separation of uranium

compounds 131. Neither for the thermal [4, 51, low pressure

electrical discharges, nor for laser induced decomposition[3],

has sufficient information been reported on the reaction pro-

ducts and their qualitative and quantitative analysis.

Some information has been found in the work of Emeleus and

Tittle [6], which was carried out in a microwave discharge,

where the discharge conditions, the product distribution and

the analytical techniques utilized, were different from tho-

se reported in the present paper.

In the present investigation we report an accurate method

for the qualitative and quantitative analysis of the products

of decomposition of SF6 which has been subjected to an RF

electrical discharge, in a quartz tube flow reactor, at a

pressure of 20 torr.

Under our experimental conditions, the qualitative analysis

mass of the discharged gases, carried out by means of IR and gas

spectrometry, has shown the presence of sulfur hexafluoride

(SF6)' sulfuryl fluoride (S02F2), thionyl fluoride (SOF2),

sulfur oxytetrafluoride (SOF4), silicon tetrafluoride (SiF4

and fluorine (F2). The presence of the oxygenated compounds

and of SiF4 in the gas mixture derives from the reaction of

1

fluorine atoms or SFx radicals with the quartz reactor walls.

The principal difficulty encountered for the analysis of this

117

type of gas mixture arises from the high reactivity and the

corrosive action of the components.

Gas mixtures of inorganic volatile fluorides have been se-

parated by gas-chromatography on 33% Kel-F on Chromosorb W [7]

and on 50% halocarbon oil on Kel-F powder m]. The mixtures ana-

lyzed, however, do not contain all the fluorides which appear

in our reaction products.

A gas mixture very similar to the one obtained under our

conditions has been analyzed by Engelbrecht et al. [9] using a

stainless steel column containing halocarbon oil on silica-gel.

The result cannot, however, be considered satisfactory both for

the pronounced tail of the SiF4 peak and for the poor reso-

lution of S02F2, SOF2 and SF4.

Padma and Vasudeva DO] have analyzed the thermal decompo-

sition products of SF6 in a quartz reactor (SFG, SF4, SOF2,

SiF4, OF*) by means of titrimetric determinations. The authors,

however, admit that this method is unsatisfactory for a com-

plete quantitative estimate of the SiF4 and OF2 content.

The quantitative analysis of fluorine has been effected by

C.W. Weber and O.H. Howard pl] by a photometric method, after

converting fluorine to chlorine in a NaCl reactor while P.C.

Nordine and D.E. Rosner [12] have utilized the reaction of

fluorine with mercury.

EXPERIMENTAL

Qualitative analysis

The discharge products, sampled from the quartz reactor by

means of an independent pumping system and collected in cali-

brated Pyrex bulbs equipped with Teflon stopcocks, were essen-

118

tially SF6, SO2F2, SOF2, SOF4, SiF4 and F2. These products,

with the exception of fluorine, have been identified by IR

spectroscopy, utilizing a Perkin-Elmer model 577 spectrome-

ter and a 8 cm cell equipped with NaCl windows. In Table I

we report the characteristic frequencies of the bands which

have been utilized for the identification of compounds and

in Fig. 1 the complete spectrum of the mixture is shown in full

lines, for two different values of pressure. The dotted line

of Fig. 1 shows the spectrum of the gas mixture after admis-

sion of water vapor into the cell. The SiF4 band at 1030 cm -1

disappears, that of SOF4 at 1380 cm -1

decreases, while the

intensities of the bands due to SO2F2 (1270; 1503 cm-') remar-

kably increase. This is the result of hydrolytic processes and

the formation of S02F2 from SOF4, according to the

SOF4 + H20 -+ S02F2 + 2HF

The presence of fluorine was detected by pouring mercury

into the sampling bulb and by observing the formation of the

fluoride.

The separation of the components of the gas mixture has

been attempted using three different gas-chromatographic co-

lumns as shown below:

No Solid support Liquid phase Loading Stainless Steel Column

Dimension Temp. ("C)

1 Silica gel Kel-FN.10 30 4m x 6mm 30

2 Porapak-Q Kel-FN.10 1 2m x 3mn 40

3 Porapak-QS ---- __ 2m x 31mn 60

TABLE

I

I.R.

characteristic

frequencies

(in

cm -')

of fluorine

compounds

sF6

SOzF2

SOF4

SOF2

SiF4

Lit.11 1

Obs

Lit.121

Obs.

Lit.131

Obs.

Lit.141

Obs.

Lit.151

Obs.

615(

s)

613(s)

848(m)

850(m)

637(s)

630(s)

748(s)

750(s)

1031(s)

1030(s)

94o(s)

945(s)

885(s)

888(s)

752(m)

760(m)

808(s)

810(s)

1269(s)

1270(s)

82l(vs)

82l(vs)

1353(s)

1330(s)

1502(s)

1503(s)

928(s)

930(s)

1383(s)

1380(s)

(1)

R.T.

Lagemann

and

E.A.

Jones,

J. Chem.

Phys.

19(15),

534

(1951).

-

(2)

W.D.

Perkin

and

M.

Kent

Wilson,

J. Chem.

Phys.

20(11),

1791

(1952).

-

(3)

F.B.

Dudley,

G.H.

Cady

and

D.F.

Eggers

Jr.,

J. Amer.

Chem.

Sot.

78,

1553

(1956).

-

(4)

R.J.

Gillespie,

E.A.

Robinson,

Can.

J. Chem.

2,

2171

(1961).

(5)

E.A.

Jones,

J.S.

Kirby-Smith,

P.J.H.

Woltz

and

A.H.

Nielsen,

J.

Chem.

Phys.

19,

242

(1951).

I.20

*---r------T 0 is 52 %-

(%) 33NVlIIVUSNVUl

121

All the analyses were carried out on a Hewlett-Packard gas chro-

matograph model 5700A, with thermoconductivity detector. The

carrier gas was helium and the gases were injected with a gas

sampling valve, after the admission of mercury into the glass

bulbs.

The first column was found to give an insufficient separa-

tion between SiF4 and SF6 and between SOFq, SO2F2 and SOF2. The

second column gave a good separation between SiF4, SF6 and SOF2,

but low resolution between SOFq and SO2F2. The best results

were obtained utilizing the third type of column and a typical

gas-chromatogram is reported in Fig. 2. The identification of

60-

46-

20-

o-

L

5 ‘!A I 1 1 L I I 1 I I

0 4 8 12 16

TIME (min)

Fig. 2 - Typical gas-chromatogram of some fluorine ComDounds.

122

the peaks was made utilizing pure samples of SFS, SiF4 and

S02F2. The last peak was identified by trapping the eluted

component directly in an evacuated infrared gas cell,which

showed the characteristic spectrum of SOF2. The assignement

of the peak of SOF4 was made by exclusion on the basis of the

IR analysis.

In most of the previously quoted investigations on SF6 de-

composition, the presence of SF4 has been reported and in many

cases the gas-chromatographic separation of SF4 from SOF2 was

difficult. In order to ascertain the presence of SF4 in

our gas mixtures, we have carried out a gas-mass spectrometric

analysis (HP model 5990/A), which has confirmed the identifica-

tion of all the G.C. peaks but excluded definitely the presence

of SF4. Table II shows the retention times, the relative abun-

dances of the parent molecule ions and the most plentiful ionic

species derived therefrom. When ionic species were sufficiently

plentiful, it was also possible to detect the presence of

33S and 34 S isotopes (1 and 5% respectively).

Quantitative analysis

After the admission of mercury into the glass bulbs to

convert fluorine in fluoride, the sample was introduced in-

to the gas-chromatograph for quantitative analysis. The

system of sample introduction is illustrated in Fig. 3. It

consists of a conventional gas sampling valve and of a va-

cuum line fitted with a pressure gauge (accuracy 1%). The

discharge gases, sampled at about 15 torr, were admitted

from the bulbs to the previously evacuated sample loop,

TABLE

II

Mass Spectrum of the

discharge

products,

obtained at 70 eV.

SUBSTANCE

(Time

retention)

sec.

TYPE OF PEAK

Mass/Charge

ratio

- Relative intensities

SiF4

SiF

SiF3

SiF4

2gSiF3

(178)

47-1.7

85-100

104-2.1

86-5.3

sF6

SF

SF4,d

SF2

sF3

SF4

sF5

(349)

51-5.3

54-1.3

70-5.4

89-32.5

108-16.0

127-100

SO2F2

SO

SF

SO2

SOF

SF2

S02F

SOF2

S02F2

(482)

48-6

51-2.4

64-6.3

67-22.3

70-2.2

83-100

86-2.3

102-86.9

SOF4

so

SF

SOF

SF2

SOF2

sF3

SOF3

(571)

48-2.6

51-1.9

67-8.9

70-2.5

86-14.7

89-7.2

105-100

SOF2

(738)

so

SF

SOF

SF2

SOF2

48-13.5

51-4.5

67-100

70-3.8

86-54.5

124

PUMPING

SYSTEM +

MANOMETER GAS SAMPLING

VALVE

+ TO COLUMN

t HELIUM

bfJ u

SAMPLE

SiF4 BULB

Fig. 3 - Scheme of gas-chromatographic inject

and after recording the gas pressure injected

ion system.

into the co-

lumn. Typical analytical conditions adopted are given in Ta-

ble III.

The carrier gas was dried in a condenser containing mo-

lecular sieves cooled with liquid nitrogen, in order to pre-

vent the fast hydrolysis of the reactive components.

TABLE III.

Operating Conditions.

Instrument

Column

Oven temperature

Detector

Current

Detector block temperature

Helium flow rate

Sample volume

Time for single analysis

Hewlett-Packard gas-chromatograph Model 5700A

and Hewlett-Packard integrator, Model 3370A

6.5 ftxl/8in, stainless steel column, Porapak-QS

60 "C

Thermal conductivity detector

270 mA

100 "C

20 cc/min.

3 ml

13 min.

125

Calibration curves of SF5, SO2F2 and SiF4 have been obtai-

ned by utilizing the corresponding pure gases (Matheson). It

was observed that successive injections of equal amounts of

SiF4 produced agradual increase in the heights of the relati-

ve peaks up to a constant maximum value. This was clearly

due to the absorption of small amounts of the gas into the co-

lumn. An injection of about 200 torr of pure SiF4 before each

set of analyses was sufficient to prevent this effect for at

least an hour.

The calibration curve of SOF2 was obtained indirectly,

after the gas-chromatographic analysis. The gas sample was

subjected to base hydrolysis and the resulting sulphites

titrated iodometrically.

The quantitative measurement of SOF4 was obtained by a

difference between the total pressure of the sampled gas

and the sum of the partial pressures of the other components.

The values thus obtained were in good agreement with those

resulting from the internal oxygen balance of Si02

'SiF4 = 'S02F2 + 3(pSOF4 + 'SOF,) (1)

The right side of equn. (1) does not contain the partial

pressure of free oxygen because it completely reacts to gi-

ve the other mentioned compounds.

Calibration plots, obtained as described above, are shown

in Fig. 4, for pressures up to 1.5 torr using a 3 ml sampling

volume. With the exception of SiF4, which shows a pronounced

curvature, the plots are linear. The uncertainty in the de-

tector calibration has been found for all gases to be f 1%

(relative error) over 0.05-10 torr pressure range and this

126

SAMPLE PRESSURE, (torr)

Fig. 4 - Calibration plots.

error is mainly due to the accuracy of the pressure gauge of

the sampling system.

The lowest detection limit reached for all the components

of the mixture was 0.05 torr, with the exception of SiF4, the

detection limit of which was 0.2 torr. These data are given

in torr for the component in the mixture, at ambient tempe-

rature, for a total pressure of 10 torr, using a 3 ml sample

loop volume.

After the gas-chromatographic analysis, the glass bulb

containing the gas mixture, the mercury fluoride and the

unreacted mercury, was evacuated from the remaining gases.

The mercury fluoride was then dissolved in distilled water

and the fluoride ion determined

thorium nitrate solution, using

sulfonate. Nitric acid has been

127

volumetrically by means of

as indicator sodium alizarin

used for acidifying the so-

lution, instead of HCl as reported by 3.M. Salsbury et al.[13],

in order to prevent the formation of calomel.

CONCLUSIONS

With the analytical technique outlined above we have been

able to study the decomposition of SF6 under R.F. discharge

at 20 torr, as a function of the power density and gas flow

rate [14]. For particular conditions of power density and

flow rate a 100% conversion is almost reached. A typical

distribution of the products, in terms of molar fraction,

is: 0.06 SF6; 0.33 SiF4; 0.27 SOF4; 0.16 S02F2; 0.10 F2;

0.07 SOF2. The presence of SiF4 as a major reaction product

is a clear indication of the high concentration of atomic

fluorine and/or SFx radicals which rapidly react with the

reactor walls. The system can therefore be profitably utili-

zed for the dry etching of the semiconductor surfaces.

The formation of SiF4 is an indication of the possibility

of gasification of silica particles to give silicon tetra-

fluoride, whose reduction with H2 under plasma conditions can

lead to solar grade silicon for the production of photovoltaic

cells.

Another possible utilization of the fluorine atoms produced

under discharge conditions could be, as already mentioned in

the introduction, the production of excited hydrogen fluoride

for laser emission. The etching of reactor walls must at all

times be avoided. Preliminary results obtained in our labora-

128

tory using an alumina tube reactor fed with mixtures of SF6

and 02, have shown the clear absence of SiF4 and the presence

of relevant amounts of fluorine, togheter with the same oxyge-

nated compounds

ACKNOWLEDGMENTS

reported above.

The authors wish to thank Mr. G. Latrofa for technical

assistence and Mr. A. Di Lorenzo of Combustion Laboratory

of C.N.R. (Naples) for helpful1 discussion.

REFERENCES

1

2

3

4

8

9

10

W. Becher and J. Massonne, Elektrotech. Z. Ausg. A,

91, 605 (1970).

C.J. Mogab, A.C. Adams, and D.L. Flamm, J.Appl. Phys.,

49(7), 3796 (1978). -

F. Brunner, T.C. Cotter, K.L. Kompa and D. Proch,

J. Chem. Phys. 67, 1547 (1977). -

K.L. Wray and E.V. Feldman, 14" Symposium (Int.) on Combusion

229 (1973).

J.F. Bott and T.A. Jacobs, J. Chem. Phys., 50, 3850 (1969).

H.J. Emeleus and B. Tittle, J. Chem. Sot., S, 1964 (1963).

R.H. Campbell and B.J. Gudzinowicz, Analytical Chem. 33, -

842 (1961).

I. Lysyj and P.R. Newton, Analytical Chem. 35, 90 (1963).

A. Engelbrecht, E. Nachbaur and E. Mayer, J. Chromatog. 15, -

228 (1964).

D.K. Padma and A.R. Vasudeva Murthy, J. Fluorine Chem. 5,

181 (1975).

129

11. C.W. Weber and O.H. Howard, Analytical Chem. 35, 1002 (1963). -

12. P.C. Nordine and D.E. Rosner, Trans. Faraday Sot. 11, 1526 (1975).

13. J.M. Salsbury, J.W. Cole, L.G. Overholser, A.R. Armstrong

and J.H. Yoe, Analytical Chem. 23, 603 (1951). -

14. G. Bruno, P. Capezzuto, F. Cramarossa and G. Latrofa,

Proceedings of the 'XI Congress0 Nazionale di Chimica Inor

ganica’ , Cosenza (Italy) September 1978.